fechter wrote:Increasing the volume by decreasing the inner radius will only help a little, as the iron between the magnets will tend to saturate first, though satruating the back iron isn't all that bad as long as there isn't any aluminum right behind it.
That was my initial thinking as well, but after considering it some more I don't think this is as much of an issue.
At lower frequencies (where inductance is not an issue), electric currents will tend to spread over all the available copper, even if it's not the most direct path. If you have a PCB with ground plane, the return currents will tend to spread over the entire plane (more concentrated on the direct path between source and load, but widely distributed). This, incidentally, is where ground loops come from. The currents tend to be spread according to the resistances of the various paths.
Since we can make a good analogy to magnetic circuits, the same principle will hold true. As the iron directly behind the magnets begins to saturate, the flux will tend to spread out more and make use of all available iron even if the path length is a little longer. It might not be quite the same effectiveness as if you put the same additional volume of iron directly behind, but I think it would still be effective enough to be worthwhile. It's impossible to simulate this 3D flux pattern in FEMM, though.